Parametrizing the Conditionally Gaussian Prior Model for Source Localization with Reference to the P20/N20 Component...

Area of Science:

• Neuroscience
• Biophysics
• Computational Biology

Background:

• Source localization of brain activity is crucial for understanding neural processes.
• Existing methods, like the conditionally Gaussian hierarchical Bayesian model (CG-HBM), have limitations in applicability and parametrization.
• Somatosensory evoked potential (SEP) and field (SEF) measurements are key for studying sensory processing.

Purpose of the Study:

• To enhance the applicability of the CG-HBM for brain activity source localization.
• To propose a robust parametrization approach for focal source scenarios.
• To achieve parametrization invariant to noise level and source space size variations.

Main Methods:

• Development of the conditionally Gaussian hierarchical Bayesian model (CG-HBM) as a superclass of inversion methods.
• Optimization of the shape parameter to minimize posterior difference between gamma and inverse gamma hyperpriors.
• Introduction of a new concept, prior-over-measurement signal-to-noise ratio, for determining the scale parameter range.

Main Results:

• The CG-HBM successfully detected the primary generator of the P20/N20 component in Brodmann area 3b.
• A robust parameter range was derived using established knowledge from Tikhonov-regularized minimum norm estimate.
• Enhanced detection of deep thalamic activity concurrent with the P20/N20 component was observed using the gamma hyperprior with an optimized shape parameter.

Conclusions:

• The proposed robust parametrization significantly improves the applicability of the CG-HBM for brain source localization.
• The new method provides accurate localization of neural activity, exemplified by the P20/N20 component.
• The CG-HBM, with optimized parameters, shows potential for detecting simultaneous deep and cortical brain activity.